Author: Ivan Bongiorni, Data Scientist at GfK; LinkedIn.

TensorFlow 2.0 Notebooks

This is a collection of my Notebooks on TensorFlow 2.0

The training of models is based on TensorFlow's eager execution method. I'll try to minimize referencese to Keras.

Summary of Contents:

• Basic feed forward stuff
• Autoencoders
• Convolutional Neural Networks
• Recurrent Neural Networks
• Applications to NLP

Contents:

Basic feed forward stuff:

• Basic classifier: implementation of a feed forward Classifier with simple, full-Batch Gradient Descent in Eager execution.

• Mini batch gradient descent: training a model with Mini Batch Gradient Descent.

• Save and restore models: how to train a model, save it, then restore it and keep training.

• Train a Neural Network with frozen layers

Autoencoders:

• Autoencoder for dimensionality reduction: implementation of a stacked Autoencoder for dimensionality reduction of datasets.

• Denoising Autoencoder (see CNN section below).

• Recurrent Autoencoder (see RNN section below).

Convolutional Neural Networks:

• Basic CNN classifier: a basic Convolutional Neural Network for multiclass classification.

• Advanced CNN classifier with custom data augmentation.

• Mixed-CNN classifier.

• Denoising Autoencoder.

Recurrent Neural Networks:

• LSTM many-to-one forecast model

• LSTM many-to-many forecast model

• Multivariate LSTM regression.

• Seq2seq models.

RNN + Natural Language Processing

• LSTM Text generator from this repository of mine.